Bayesian updating
How should a trader change their view when new market data arrives, and how much should one week move that view?
In plain words
Bayesian updating is the practice of revising a belief by a measured amount when new evidence arrives, rather than keeping an old view unchanged or discarding it entirely. The size of the revision depends on how informative the new data is and how confident the prior belief was.
Why it matters
Traders carry views about direction, volatility and risk from one week to the next. Treating each new datapoint as decisive causes whipsawing; ignoring it causes stale positioning. A week of mixed, low-conviction evidence should move a view only slightly, while a strong, high-participation week justifies a larger revision.
In a real week: BTC, week of 20 March 2023
In the week of 20 March 2023, BTC closed 27968.05, a return of -0.02% against an open of 27972.87, after prior weeks of -2.96%, -4.77%, -1.93% and +27.16%. Volume of 40.09 billion dollars was 0.86 of its prior 12-week average, and the close sat in the middle third of the week range at 60.3%. This week ranked in the 46.0 return percentile across 266 weeks of history: flat, mixed evidence warranting only a small revision in either direction.
A common mistake
Revising a view as much as the week's headline swing suggests. A wide range or a sharp single hour does not by itself supply strong evidence about the trend, especially when volume and the weekly close tell a quieter story.